•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`Diversity Combining within Viterbi
`
`Memo
`Subject:
`
`Author:
`
`Ernst Eberlein
`
`Date:
`
`10/26/98
`
`1 Scope
`This memo shall give additional information to the new proposed diversity combining
`scheme.
`
`2 Description of the diversity combining method
`According to the following block diagram the AMRC OARS system can be considered as
`transmission system with one channel.
`
`Bit steam
`source
`
`Forward
`Error
`Correction
`
`De(cid:173)
`multiplex
`
`Satellite
`~ - - - - - #1
`
`Delay
`
`Satellite
`#2
`
`Channel 1----- Receiver
`
`According to the code rate of¾ (Convolutional code only) for 3 information bits 4 channel
`bits are transmitted over each satellite. Using two satellites 8 channel bits are transmitted
`for 3 information bits. Therefore the system can be considered as a system with code rate
`3/8.
`
`According to the literature and system simulation results the following Eb/No performance
`can be assumed. For QPSK the C/N (=Power of transmitted signal/Noise power within
`effective bandwidth) can be calculated by:
`
`% = E7M0 +lO*log(R)+3dB
`
`( C/N and Et,/No values in dB)
`
`R is the code rate.
`
`Code Rate R (convolutional code +
`Reed-Solomon Code)
`3/4 * (223/255) = 0.66
`1/2 * (223/255) = 0.44
`3/8 * (223/255) = 0.33
`
`Et,/No [dB]
`
`C/N [dB]
`
`3.7
`2.7
`aoo. 2.4
`
`4.9
`2.1
`0.6
`
`Diversity_combining_memo2.doc
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 1
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`•-===
`•-===
`lml
`Fraunhofer lnstitut
`The required C/N value applies also to a system where the
`lntegrierte Schaltungen
`bitstream is de-multiplexed to two stream and transmitted using 2 QPSK modulators. The
`overall transmit power is
`C = Csat 1 + C Sat 2
`
`The noise power is
`
`Assuming the signal power is the identical (=best case for the combining method) for
`satellite 1 and 2 Csa11 = Csa12. With N1 = N2 (the effective bandwidth is identical for both
`signals the following equation applies:
`
`Assuming the available signal power (=C) and QPSK symbol rate is kept the combining
`method can give an gain of 4.3dB compared to the required C/N if one signal is decoded
`only. This gain is equivalent to the gain relative to switching combining for the scenario
`available C5a11/N = CsadN. It is assumed that for other scenarios the gain is lower. At least
`for the scenarios Csa11/N or CsadN is greater then 4.9dB no gain is required. The output
`signal is error free in any case. The overall gain of the scheme depend on the probability of
`the scenario.
`In other words: It is possible to receive the signal down to an C/N of 0.6dB (theoretical value
`not including implementation loss). The equivalent required C/No = 62.7 dBHz (not including
`implementation loss) for each satellite. If only one satellite signal is available the required
`C/No is 67 dBHz1 (not including implementation loss).
`
`The implementation requires a convolution al coder with code rate 1 /3. The output of the
`convolutional encoder is punctured to a code rate of 3/8 by not transmitting 1 channel bit
`out of 9. The output of the convolutional encoder and puncturing unit is demultiplexed. 4
`bits out of 8 are transmitted over satellite 1. The other 4 bits are transmitted over satellite 2.
`A delay can be inserted for one signal. Optional additional time interleaver can be used. A
`simplified block diagram (not including the details of the TOM bitstream structure) is shown
`in the next figure.
`
`.I
`I
`
`----.i
`
`gl
`
`g2
`
`g3
`
`I
`I
`
`I
`
`I
`
`Convolutional coder
`
`Puncturing
`unit
`
`Parallel to
`serial and
`De-
`multiplex
`to 2
`bitstreams
`
`~
`
`Delay
`(e.g. 4
`seconds)
`
`-------
`
`-----+
`
`The polynomial g1, g2 and g3 describe the shift registers and modulo 2 adders which
`generates the convolutional code of code rate 1/3. One bit out of 9 is removed by the
`
`1 For the WorldSpace L-Band system the required C/NO is 64. 7dBHz. The value of 67dBHz used for
`many link budget calculations includes a implementation loss of 2.3dB.
`
`Diversity_combining_memo2.doc
`
`2
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 2
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`puncturing unit. The remaining 8 bits are de-multiplexed to the
`two output bitstreams according to the scheme shown in the next figure.
`
`•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`Diversity_combining_memo2.doc
`
`3
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 3
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`Input bit sequence:
`
`- - -+ time
`
`After convolutional encoder
`
`I~ In ~1111
`
`I
`
`E = Bit transmitted over early satellite
`L = Bit transmitted over late satellite
`X = not transmitted (punctured) bit
`
`The proposed polynomials are
`g 1 = 1001111 (same as for current spec.)
`g2 = 1101101 (same as for current spec.)
`93=1010111
`
`The receiver requires one Viterbi decoder only. A simplified block diagram of the receiver is
`shown in the next figure:
`
`QPSK
`Demod
`
`TDM
`Demux
`
`Delay
`(Memory)
`
`QPSK
`Demod
`
`TDM
`Demux
`
`De-
`Puncture
`
`Viterbi
`Decoder
`
`Reed-
`Solomon
`
`The optimal combining according to the signal quality of the two signals is automatically
`performed by the Viterbi decoder. The Viterbi decoding performs maximum likelihood
`decoding using the channel state information (="metric"). The algorithm used for the metric
`calculation is TBD. Algorithms known for Rician and Rayleigh channels can be adapted to
`the AMRC OARS system.
`
`If only one signal is available the input of the Viterbi decoder is considered as convolutional
`code of code rate 1/3 punctured to a code rate of¾. The equivalent puncturing scheme is:
`
`For the early satellite:
`1 1 1
`100
`000
`
`For the late satellite
`000
`001
`1 1 1
`
`Diversity_combining_memo2.doc
`
`4
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 4
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`3 Impact to the system design
`
`3.1 Waveform specification
`The available waveform specification proposes the same convolutional code for both
`satellites and "staggered puncturing" resulting in a TOM bitstream different for the early and
`the late satellite. The convolutional encoder and the staggered puncturing shall be replaced
`by the definition of the new convolutional code (g3 shall be added) and the description of
`the puncturing and "de-multiplex" of the bitstream. Using g 1 and g2 for satellite #1 and g2
`and g3 for satellite #2 together with modified puncturing scheme would be an equivalent
`specification.
`No changes for the terrestrial waveform are required. The reformating of the bitstream
`received from the early satellite will be unmodified (if one signal is decoded only the
`convolutional code can be considered as code rate ½ punctured to¾ also).
`
`3.2 System analysis/simulation
`A simulation setup containing the following building blocks shall be developed and tested:
`- convolutional encoder according to the proposal
`- puncturing and demultiplex
`- 2* TOM bitstream multiplex
`- delay for one signal
`- 2* QPSK modulator
`- Channel model validated with SV2 data
`- 2* AGC
`- 2* QPSK demodulator
`- delay for one signal
`- multiplex of the two softquantized bitstreams
`- de-puncturing
`- Viterbi decoder with metric calculation adapted to the requirements (if necessary)
`
`To avoid problems with the simulation setup using long delays for an initial analysis it is
`sufficient to replace the delay of 4 seconds by a delay greater to the coherence time of the
`channel ''tree coverage". Values in the range of app. 1 OOms shall be sufficient.
`Assuming the channel models and the AGC concept can be taken from the setups used for
`MRC and/or post-FEC the main effort results from the development of the modified Viterbi.
`
`3.3 Schedule impact
`The goal of the schedule is a frozen specification (waveform and chipset) by January 15th ,
`1999. To meet this milestone preliminary results shall be available beginning of December
`to adapt the draft specification, which will be developed in parallel to the SV1/SV2 data
`analysis, to the results of the system validation.
`Proposal:
`1. If the data available by Oct 30th indicates that post-Viterbi and post-Reed-Solomon
`decoding are not sufficient or an additional gain of 2-4 dB will improve the service
`availability significant the combining methods before/within Viterbi decoding shall be kept as
`option for the baseline receiver. The decision shall be postponed until end of November
`2. The additional analysis shall be performed until end of November. The additional analysis
`shall include:
`• Analysis of "Code rate 1/3 approach" by system simulation
`• Development of draft chipset spec. based on the assumption
`
`Diversity_combining_memo2.doc
`
`5
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 5
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`• external memory will be used for the delay
`• Viterbi decoder for code rate 1/3
`• Metric calculation is based on the QPSK output only (field strength and
`AGC value is not used by the metric calculation -> memory for storing
`broadcast channel data will be sufficient
`• Analysis of degradation cause by 2bit soft-quantization (compared to 4 bit as
`proposed for the baseline) -> 50% memory only sufficient?
`
`The results shall be compared to the MRC algorithm.
`
`3.4 Receiver Cost
`The cost of the new proposal is the same or less then the MRC algorithm. The method is
`only feasible if external memory is used for the delay. The required memory size is
`806kbit/second = 3.5 Mbit for 4.32second (= 8.25kbps/slot * 16slots *1/code_rate *
`4bit/softquantized bit* 4.32seconds. The code rate is¾* (223/255)).
`
`Diversity_combining_memo2.doc
`
`6
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 6
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`Annex - New proposal on diversity combining ''within Viterbi
`decoder''
`
`•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`The diversity concept can be considered as repetition code. Generating the two TOM
`bitstreams can be considered as convolutional code with code rate 3/8 which is derived
`from a mother code with code rate 1/4 with puncturing to 3/8. In the current design the
`polynomials for the convolutional encoder for the early and late satellite are identical. The
`puncturing scheme is different ("staggered puncturing"). Therefore we can consider the
`convolutional code as a coder with 4 shift registers and puncturing. The output of two shift
`registers are transmitted over the early satellite. The output of the other two shift registers is
`transmitted over the late satellite. For each useful bit 4 channel bits are generated (3 bits
`give 12bits = 2 streams, 3*2 bits). 4 bits out of 12 channel bits are punctured resulting in a
`remaining code rate of 3/8. 4 bits are transmitted over early satellite. The other 4 bits are
`transmitted are over the late satellite.
`Up to now we consider that the two streams are decoded by two Viterbi decoders or
`combined using MRC and the send to one Viterbi decoder.
`It is possible to decode the two streams by one Viterbi decoder without MRC combining
`before the Viterbi decoder. On the receiving side the bits from the early and late satellite are
`multiplexed to one stream (the signal of the early signal must be delayed of course). The
`punctured bits are re-inserted. The resulting stream is decoded by an Viterbi decoder for
`code rate ¼. Optional a mother code of code rate 1/3 can be used also.
`
`The following diagram shows the concept for mother code with code rate 1/3.
`
`Input bit sequence:
`
`After convolutional encoder
`
`E = Bit transmitted over early satellite
`L = Bit transmitted over late satellite
`X = not transmitted (punctured) bit
`
`In the receiver the bits received from the early satellite and late satellite are multiplexed
`together to generate the Viterbi-Decoder input. Punctured bits are replaced by "unknown"
`bits. Then a standard Viterbi decoder for code rate 1/3 can be used.
`If only one satellite is available only all bits transmitted over the missing satellite are
`replaced by the "unknown" bits. This is equivalent to puncturing of code rate 1/3 = 3/9 down
`to¾.
`
`Theoretical Performance:
`The required Eb/NO for a system with concatenated FEC using convolutional code of code
`rate 3/8 and Reed-Solomon code of type (223,255) is app. 2.4dB (0.3 dB better then ½
`+(223,255). This value applies to the best case (both signals have equal level). This is
`equivalent to an C/N of 0.6 dB. For MRC the theoretical best case performance is 2dB.
`If one signal is used only the required C/N is SdB.
`
`Diversity_combining_memo2.doc
`
`7
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 7
`Sirius v Fraunhofer
`IPR2018-00690
`
`

`

`•-===
`•-===
`lml
`Fraunhofer lnstitut
`lntegrierte Schaltungen
`
`Therefore the method offers in theory a performance increase of
`1.4 dB (compared to MRC) or 4.4 dB (compared to switching combining).
`Please note: This gain applies only to the scenario both signals have equal level and are
`close to the threshold. If one signal is above the threshold no gain is required (the output is
`error free in any case due to the brickwall characteristic of the concatenated FEC scheme).
`If the signal level difference is high no gain is expected.
`
`Impact to the System design:
`• A Viterbi decoder for code rate 1/3 shall be proposed
`• The signals transmitted over the early and late satellite are different (as planned in any
`case, only the "staggered puncturing" scheme changes).
`• A method detecting "signal not available" shall be developed and tested
`• Validation of the performance by system simulation.
`
`The straight forward design simplifies the validation effort. A simulation setup with simple
`channel models (e.g. variation of the fieldstrength only+ arbitrary phase after a dropout)
`shall be sufficient.
`It is proposed to test the algorithm by COSSAP simulation using the available QPSK
`demodulator, TOM multiplexing/demultiplexing modules with modified convolutional
`encoder/decoder, a delay line to simulate the early/late satellite delay and a channel model
`derived from SV2 data (e.g. fieldstrength and phase measurements).
`
`Diversity_combining_memo2.doc
`
`8
`
`Ebl { DATE }
`
`Fraunhofer Ex 2051-p 8
`Sirius v Fraunhofer
`IPR2018-00690
`
`